submit depthwise_conv_op and test

fix #318 program add split

Init tensor memory in executor_for_test

commit io files

modify operator interface for printing tensor array

fix #311 restore test for single input and output

fix #309 

refine unit test

fix #305

fix #303 add fushion fc

update Gemm with implementation of 'C = alpha * A * B + beta * C'

fix #298 optimize some ops

fix #296 add resize output tensor

fix ubuntu compile